blob: 49b3ea615dc5fa991d921667f44172da02331a55 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/ipc/util.c
3 * Copyright (C) 1992 Krishna Balasubramanian
4 *
5 * Sep 1997 - Call suser() last after "normal" permission checks so we
6 * get BSD style process accounting right.
7 * Occurs in several places in the IPC code.
8 * Chris Evans, <chris@ferret.lmh.ox.ac.uk>
9 * Nov 1999 - ipc helper functions, unified SMP locking
Christian Kujau624dffc2006-01-15 02:43:54 +010010 * Manfred Spraul <manfred@colorfullife.com>
Linus Torvalds1da177e2005-04-16 15:20:36 -070011 * Oct 2002 - One lock per IPC id. RCU ipc_free for lock-free grow_ary().
12 * Mingming Cao <cmm@us.ibm.com>
Steve Grubb073115d2006-04-02 17:07:33 -040013 * Mar 2006 - support for audit of ipc object properties
14 * Dustin Kirkland <dustin.kirkland@us.ibm.com>
Kirill Korotaev73ea4132006-10-02 02:18:20 -070015 * Jun 2006 - namespaces ssupport
16 * OpenVZ, SWsoft Inc.
17 * Pavel Emelianov <xemul@openvz.org>
Linus Torvalds1da177e2005-04-16 15:20:36 -070018 */
19
Linus Torvalds1da177e2005-04-16 15:20:36 -070020#include <linux/mm.h>
21#include <linux/shm.h>
22#include <linux/init.h>
23#include <linux/msg.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070024#include <linux/vmalloc.h>
25#include <linux/slab.h>
Randy.Dunlapc59ede72006-01-11 12:17:46 -080026#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <linux/highuid.h>
28#include <linux/security.h>
29#include <linux/rcupdate.h>
30#include <linux/workqueue.h>
Mike Waychisonae781772005-09-06 15:17:09 -070031#include <linux/seq_file.h>
32#include <linux/proc_fs.h>
Steve Grubb073115d2006-04-02 17:07:33 -040033#include <linux/audit.h>
Kirill Korotaev73ea4132006-10-02 02:18:20 -070034#include <linux/nsproxy.h>
Nadia Derbey3e148c72007-10-18 23:40:54 -070035#include <linux/rwsem.h>
Nadia Derbeyb6b337a2008-04-29 01:00:42 -070036#include <linux/memory.h>
Pavel Emelyanovae5e1b22008-02-08 04:18:22 -080037#include <linux/ipc_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070038
39#include <asm/unistd.h>
40
41#include "util.h"
42
Mike Waychisonae781772005-09-06 15:17:09 -070043struct ipc_proc_iface {
44 const char *path;
45 const char *header;
Kirill Korotaev73ea4132006-10-02 02:18:20 -070046 int ids;
Mike Waychisonae781772005-09-06 15:17:09 -070047 int (*show)(struct seq_file *, void *);
48};
49
Kirill Korotaev73ea4132006-10-02 02:18:20 -070050struct ipc_namespace init_ipc_ns = {
51 .kref = {
52 .refcount = ATOMIC_INIT(2),
53 },
54};
55
Nadia Derbey4d89dc62008-04-29 01:00:40 -070056atomic_t nr_ipc_ns = ATOMIC_INIT(1);
57
58
Nadia Derbeyb6b337a2008-04-29 01:00:42 -070059#ifdef CONFIG_MEMORY_HOTPLUG
60
Nadia Derbey424450c2008-04-29 01:00:43 -070061static void ipc_memory_notifier(struct work_struct *work)
62{
63 ipcns_notify(IPCNS_MEMCHANGED);
64}
65
66static DECLARE_WORK(ipc_memory_wq, ipc_memory_notifier);
67
68
Nadia Derbeyb6b337a2008-04-29 01:00:42 -070069static int ipc_memory_callback(struct notifier_block *self,
70 unsigned long action, void *arg)
71{
72 switch (action) {
73 case MEM_ONLINE: /* memory successfully brought online */
74 case MEM_OFFLINE: /* or offline: it's time to recompute msgmni */
75 /*
76 * This is done by invoking the ipcns notifier chain with the
77 * IPC_MEMCHANGED event.
Nadia Derbey424450c2008-04-29 01:00:43 -070078 * In order not to keep the lock on the hotplug memory chain
79 * for too long, queue a work item that will, when waken up,
80 * activate the ipcns notification chain.
81 * No need to keep several ipc work items on the queue.
Nadia Derbeyb6b337a2008-04-29 01:00:42 -070082 */
Nadia Derbey424450c2008-04-29 01:00:43 -070083 if (!work_pending(&ipc_memory_wq))
84 schedule_work(&ipc_memory_wq);
Nadia Derbeyb6b337a2008-04-29 01:00:42 -070085 break;
86 case MEM_GOING_ONLINE:
87 case MEM_GOING_OFFLINE:
88 case MEM_CANCEL_ONLINE:
89 case MEM_CANCEL_OFFLINE:
90 default:
91 break;
92 }
93
94 return NOTIFY_OK;
95}
96
97#endif /* CONFIG_MEMORY_HOTPLUG */
98
Linus Torvalds1da177e2005-04-16 15:20:36 -070099/**
100 * ipc_init - initialise IPC subsystem
101 *
102 * The various system5 IPC resources (semaphores, messages and shared
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800103 * memory) are initialised
Nadia Derbeyb6b337a2008-04-29 01:00:42 -0700104 * A callback routine is registered into the memory hotplug notifier
105 * chain: since msgmni scales to lowmem this callback routine will be
106 * called upon successful memory add / remove to recompute msmgni.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107 */
108
109static int __init ipc_init(void)
110{
111 sem_init();
112 msg_init();
113 shm_init();
Nadia Derbeyb6b337a2008-04-29 01:00:42 -0700114 hotplug_memory_notifier(ipc_memory_callback, IPC_CALLBACK_PRI);
115 register_ipcns_notifier(&init_ipc_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116 return 0;
117}
118__initcall(ipc_init);
119
120/**
121 * ipc_init_ids - initialise IPC identifiers
122 * @ids: Identifier set
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123 *
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700124 * Set up the sequence range to use for the ipc identifier range (limited
125 * below IPCMNI) then initialise the ids idr.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126 */
127
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700128void ipc_init_ids(struct ipc_ids *ids)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129{
Nadia Derbey3e148c72007-10-18 23:40:54 -0700130 init_rwsem(&ids->rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132 ids->in_use = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133 ids->seq = 0;
134 {
135 int seq_limit = INT_MAX/SEQ_MULTIPLIER;
Zhang, Yanmin44f564a2008-04-29 01:00:55 -0700136 if (seq_limit > USHORT_MAX)
137 ids->seq_max = USHORT_MAX;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700138 else
139 ids->seq_max = seq_limit;
140 }
141
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700142 idr_init(&ids->ipcs_idr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143}
144
Mike Waychisonae781772005-09-06 15:17:09 -0700145#ifdef CONFIG_PROC_FS
Arjan van de Ven9a321442007-02-12 00:55:35 -0800146static const struct file_operations sysvipc_proc_fops;
Mike Waychisonae781772005-09-06 15:17:09 -0700147/**
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800148 * ipc_init_proc_interface - Create a proc interface for sysipc types using a seq_file interface.
Mike Waychisonae781772005-09-06 15:17:09 -0700149 * @path: Path in procfs
150 * @header: Banner to be printed at the beginning of the file.
151 * @ids: ipc id table to iterate.
152 * @show: show routine.
153 */
154void __init ipc_init_proc_interface(const char *path, const char *header,
Kirill Korotaev73ea4132006-10-02 02:18:20 -0700155 int ids, int (*show)(struct seq_file *, void *))
Mike Waychisonae781772005-09-06 15:17:09 -0700156{
157 struct proc_dir_entry *pde;
158 struct ipc_proc_iface *iface;
159
160 iface = kmalloc(sizeof(*iface), GFP_KERNEL);
161 if (!iface)
162 return;
163 iface->path = path;
164 iface->header = header;
165 iface->ids = ids;
166 iface->show = show;
167
Denis V. Lunev6a6375d2008-04-29 01:02:12 -0700168 pde = proc_create_data(path,
169 S_IRUGO, /* world readable */
170 NULL, /* parent dir */
171 &sysvipc_proc_fops,
172 iface);
173 if (!pde) {
Mike Waychisonae781772005-09-06 15:17:09 -0700174 kfree(iface);
175 }
176}
177#endif
178
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179/**
180 * ipc_findkey - find a key in an ipc identifier set
181 * @ids: Identifier set
182 * @key: The key to find
183 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700184 * Requires ipc_ids.rw_mutex locked.
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700185 * Returns the LOCKED pointer to the ipc structure if found or NULL
186 * if not.
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700187 * If key is found ipc points to the owning ipc structure
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188 */
189
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700190static struct kern_ipc_perm *ipc_findkey(struct ipc_ids *ids, key_t key)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700192 struct kern_ipc_perm *ipc;
193 int next_id;
194 int total;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700196 for (total = 0, next_id = 0; total < ids->in_use; next_id++) {
197 ipc = idr_find(&ids->ipcs_idr, next_id);
198
199 if (ipc == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200 continue;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700201
202 if (ipc->key != key) {
203 total++;
204 continue;
205 }
206
207 ipc_lock_by_ptr(ipc);
208 return ipc;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209 }
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700210
211 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212}
213
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700214/**
215 * ipc_get_maxid - get the last assigned id
216 * @ids: IPC identifier set
217 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700218 * Called with ipc_ids.rw_mutex held.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219 */
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700220
221int ipc_get_maxid(struct ipc_ids *ids)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700223 struct kern_ipc_perm *ipc;
224 int max_id = -1;
225 int total, id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700227 if (ids->in_use == 0)
228 return -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700230 if (ids->in_use == IPCMNI)
231 return IPCMNI - 1;
232
233 /* Look for the last assigned id */
234 total = 0;
235 for (id = 0; id < IPCMNI && total < ids->in_use; id++) {
236 ipc = idr_find(&ids->ipcs_idr, id);
237 if (ipc != NULL) {
238 max_id = id;
239 total++;
240 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241 }
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700242 return max_id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243}
244
245/**
246 * ipc_addid - add an IPC identifier
247 * @ids: IPC identifier set
248 * @new: new IPC permission set
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700249 * @size: limit for the number of used ids
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250 *
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700251 * Add an entry 'new' to the IPC ids idr. The permissions object is
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 * initialised and the first free entry is set up and the id assigned
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700253 * is returned. The 'new' entry is returned in a locked state on success.
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700254 * On failure the entry is not locked and a negative err-code is returned.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700256 * Called with ipc_ids.rw_mutex held as a writer.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257 */
258
259int ipc_addid(struct ipc_ids* ids, struct kern_ipc_perm* new, int size)
260{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700261 int id, err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700263 if (size > IPCMNI)
264 size = IPCMNI;
265
266 if (ids->in_use >= size)
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700267 return -ENOSPC;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700268
269 err = idr_get_new(&ids->ipcs_idr, new, &id);
270 if (err)
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700271 return err;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700272
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273 ids->in_use++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274
275 new->cuid = new->uid = current->euid;
276 new->gid = new->cgid = current->egid;
277
278 new->seq = ids->seq++;
279 if(ids->seq > ids->seq_max)
280 ids->seq = 0;
281
Pierre Peiffer48dea402008-04-29 01:00:35 -0700282 new->id = ipc_buildid(id, new->seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283 spin_lock_init(&new->lock);
284 new->deleted = 0;
285 rcu_read_lock();
286 spin_lock(&new->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287 return id;
288}
289
290/**
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700291 * ipcget_new - create a new ipc object
292 * @ns: namespace
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700293 * @ids: IPC identifer set
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700294 * @ops: the actual creation routine to call
295 * @params: its parameters
296 *
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700297 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
298 * when the key is IPC_PRIVATE.
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700299 */
Pavel Emelyanovb2d75cd2008-02-08 04:18:54 -0800300static int ipcget_new(struct ipc_namespace *ns, struct ipc_ids *ids,
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700301 struct ipc_ops *ops, struct ipc_params *params)
302{
303 int err;
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700304retry:
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700305 err = idr_pre_get(&ids->ipcs_idr, GFP_KERNEL);
306
307 if (!err)
308 return -ENOMEM;
309
Nadia Derbey3e148c72007-10-18 23:40:54 -0700310 down_write(&ids->rw_mutex);
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700311 err = ops->getnew(ns, params);
Nadia Derbey3e148c72007-10-18 23:40:54 -0700312 up_write(&ids->rw_mutex);
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700313
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700314 if (err == -EAGAIN)
315 goto retry;
316
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700317 return err;
318}
319
320/**
321 * ipc_check_perms - check security and permissions for an IPC
322 * @ipcp: ipc permission set
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700323 * @ops: the actual security routine to call
324 * @params: its parameters
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700325 *
326 * This routine is called by sys_msgget(), sys_semget() and sys_shmget()
327 * when the key is not IPC_PRIVATE and that key already exists in the
328 * ids IDR.
329 *
330 * On success, the IPC id is returned.
331 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700332 * It is called with ipc_ids.rw_mutex and ipcp->lock held.
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700333 */
334static int ipc_check_perms(struct kern_ipc_perm *ipcp, struct ipc_ops *ops,
335 struct ipc_params *params)
336{
337 int err;
338
339 if (ipcperms(ipcp, params->flg))
340 err = -EACCES;
341 else {
342 err = ops->associate(ipcp, params->flg);
343 if (!err)
344 err = ipcp->id;
345 }
346
347 return err;
348}
349
350/**
351 * ipcget_public - get an ipc object or create a new one
352 * @ns: namespace
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700353 * @ids: IPC identifer set
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700354 * @ops: the actual creation routine to call
355 * @params: its parameters
356 *
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700357 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
358 * when the key is not IPC_PRIVATE.
359 * It adds a new entry if the key is not found and does some permission
360 * / security checkings if the key is found.
361 *
362 * On success, the ipc id is returned.
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700363 */
Pavel Emelyanovb2d75cd2008-02-08 04:18:54 -0800364static int ipcget_public(struct ipc_namespace *ns, struct ipc_ids *ids,
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700365 struct ipc_ops *ops, struct ipc_params *params)
366{
367 struct kern_ipc_perm *ipcp;
368 int flg = params->flg;
369 int err;
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700370retry:
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700371 err = idr_pre_get(&ids->ipcs_idr, GFP_KERNEL);
372
Nadia Derbey3e148c72007-10-18 23:40:54 -0700373 /*
374 * Take the lock as a writer since we are potentially going to add
375 * a new entry + read locks are not "upgradable"
376 */
377 down_write(&ids->rw_mutex);
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700378 ipcp = ipc_findkey(ids, params->key);
379 if (ipcp == NULL) {
380 /* key not used */
381 if (!(flg & IPC_CREAT))
382 err = -ENOENT;
383 else if (!err)
384 err = -ENOMEM;
385 else
386 err = ops->getnew(ns, params);
387 } else {
388 /* ipc object has been locked by ipc_findkey() */
389
390 if (flg & IPC_CREAT && flg & IPC_EXCL)
391 err = -EEXIST;
392 else {
393 err = 0;
394 if (ops->more_checks)
395 err = ops->more_checks(ipcp, params);
396 if (!err)
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700397 /*
398 * ipc_check_perms returns the IPC id on
399 * success
400 */
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700401 err = ipc_check_perms(ipcp, ops, params);
402 }
403 ipc_unlock(ipcp);
404 }
Nadia Derbey3e148c72007-10-18 23:40:54 -0700405 up_write(&ids->rw_mutex);
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700406
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700407 if (err == -EAGAIN)
408 goto retry;
409
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700410 return err;
411}
412
413
414/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 * ipc_rmid - remove an IPC identifier
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700416 * @ids: IPC identifier set
417 * @ipcp: ipc perm structure containing the identifier to remove
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700419 * ipc_ids.rw_mutex (as a writer) and the spinlock for this ID are held
420 * before this function is called, and remain locked on the exit.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 */
422
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700423void ipc_rmid(struct ipc_ids *ids, struct kern_ipc_perm *ipcp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424{
Nadia Derbeyce621f52007-10-18 23:40:52 -0700425 int lid = ipcid_to_idx(ipcp->id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700427 idr_remove(&ids->ipcs_idr, lid);
428
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429 ids->in_use--;
430
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700431 ipcp->deleted = 1;
432
433 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434}
435
436/**
437 * ipc_alloc - allocate ipc space
438 * @size: size desired
439 *
440 * Allocate memory from the appropriate pools and return a pointer to it.
441 * NULL is returned if the allocation fails
442 */
443
444void* ipc_alloc(int size)
445{
446 void* out;
447 if(size > PAGE_SIZE)
448 out = vmalloc(size);
449 else
450 out = kmalloc(size, GFP_KERNEL);
451 return out;
452}
453
454/**
455 * ipc_free - free ipc space
456 * @ptr: pointer returned by ipc_alloc
457 * @size: size of block
458 *
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800459 * Free a block created with ipc_alloc(). The caller must know the size
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460 * used in the allocation call.
461 */
462
463void ipc_free(void* ptr, int size)
464{
465 if(size > PAGE_SIZE)
466 vfree(ptr);
467 else
468 kfree(ptr);
469}
470
471/*
472 * rcu allocations:
473 * There are three headers that are prepended to the actual allocation:
474 * - during use: ipc_rcu_hdr.
475 * - during the rcu grace period: ipc_rcu_grace.
476 * - [only if vmalloc]: ipc_rcu_sched.
477 * Their lifetime doesn't overlap, thus the headers share the same memory.
478 * Unlike a normal union, they are right-aligned, thus some container_of
479 * forward/backward casting is necessary:
480 */
481struct ipc_rcu_hdr
482{
483 int refcount;
484 int is_vmalloc;
485 void *data[0];
486};
487
488
489struct ipc_rcu_grace
490{
491 struct rcu_head rcu;
492 /* "void *" makes sure alignment of following data is sane. */
493 void *data[0];
494};
495
496struct ipc_rcu_sched
497{
498 struct work_struct work;
499 /* "void *" makes sure alignment of following data is sane. */
500 void *data[0];
501};
502
503#define HDRLEN_KMALLOC (sizeof(struct ipc_rcu_grace) > sizeof(struct ipc_rcu_hdr) ? \
504 sizeof(struct ipc_rcu_grace) : sizeof(struct ipc_rcu_hdr))
505#define HDRLEN_VMALLOC (sizeof(struct ipc_rcu_sched) > HDRLEN_KMALLOC ? \
506 sizeof(struct ipc_rcu_sched) : HDRLEN_KMALLOC)
507
508static inline int rcu_use_vmalloc(int size)
509{
510 /* Too big for a single page? */
511 if (HDRLEN_KMALLOC + size > PAGE_SIZE)
512 return 1;
513 return 0;
514}
515
516/**
517 * ipc_rcu_alloc - allocate ipc and rcu space
518 * @size: size desired
519 *
520 * Allocate memory for the rcu header structure + the object.
521 * Returns the pointer to the object.
522 * NULL is returned if the allocation fails.
523 */
524
525void* ipc_rcu_alloc(int size)
526{
527 void* out;
528 /*
529 * We prepend the allocation with the rcu struct, and
530 * workqueue if necessary (for vmalloc).
531 */
532 if (rcu_use_vmalloc(size)) {
533 out = vmalloc(HDRLEN_VMALLOC + size);
534 if (out) {
535 out += HDRLEN_VMALLOC;
536 container_of(out, struct ipc_rcu_hdr, data)->is_vmalloc = 1;
537 container_of(out, struct ipc_rcu_hdr, data)->refcount = 1;
538 }
539 } else {
540 out = kmalloc(HDRLEN_KMALLOC + size, GFP_KERNEL);
541 if (out) {
542 out += HDRLEN_KMALLOC;
543 container_of(out, struct ipc_rcu_hdr, data)->is_vmalloc = 0;
544 container_of(out, struct ipc_rcu_hdr, data)->refcount = 1;
545 }
546 }
547
548 return out;
549}
550
551void ipc_rcu_getref(void *ptr)
552{
553 container_of(ptr, struct ipc_rcu_hdr, data)->refcount++;
554}
555
David Howells65f27f32006-11-22 14:55:48 +0000556static void ipc_do_vfree(struct work_struct *work)
557{
558 vfree(container_of(work, struct ipc_rcu_sched, work));
559}
560
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561/**
Randy Dunlap1e5d5332005-11-07 01:01:06 -0800562 * ipc_schedule_free - free ipc + rcu space
563 * @head: RCU callback structure for queued work
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 *
565 * Since RCU callback function is called in bh,
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800566 * we need to defer the vfree to schedule_work().
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 */
568static void ipc_schedule_free(struct rcu_head *head)
569{
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700570 struct ipc_rcu_grace *grace;
571 struct ipc_rcu_sched *sched;
572
573 grace = container_of(head, struct ipc_rcu_grace, rcu);
574 sched = container_of(&(grace->data[0]), struct ipc_rcu_sched,
575 data[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576
David Howells65f27f32006-11-22 14:55:48 +0000577 INIT_WORK(&sched->work, ipc_do_vfree);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578 schedule_work(&sched->work);
579}
580
581/**
Randy Dunlap1e5d5332005-11-07 01:01:06 -0800582 * ipc_immediate_free - free ipc + rcu space
583 * @head: RCU callback structure that contains pointer to be freed
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584 *
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800585 * Free from the RCU callback context.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586 */
587static void ipc_immediate_free(struct rcu_head *head)
588{
589 struct ipc_rcu_grace *free =
590 container_of(head, struct ipc_rcu_grace, rcu);
591 kfree(free);
592}
593
594void ipc_rcu_putref(void *ptr)
595{
596 if (--container_of(ptr, struct ipc_rcu_hdr, data)->refcount > 0)
597 return;
598
599 if (container_of(ptr, struct ipc_rcu_hdr, data)->is_vmalloc) {
600 call_rcu(&container_of(ptr, struct ipc_rcu_grace, data)->rcu,
601 ipc_schedule_free);
602 } else {
603 call_rcu(&container_of(ptr, struct ipc_rcu_grace, data)->rcu,
604 ipc_immediate_free);
605 }
606}
607
608/**
609 * ipcperms - check IPC permissions
610 * @ipcp: IPC permission set
611 * @flag: desired permission set.
612 *
613 * Check user, group, other permissions for access
614 * to ipc resources. return 0 if allowed
615 */
616
617int ipcperms (struct kern_ipc_perm *ipcp, short flag)
618{ /* flag will most probably be 0 or S_...UGO from <linux/stat.h> */
Steve Grubb073115d2006-04-02 17:07:33 -0400619 int requested_mode, granted_mode, err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620
Steve Grubb073115d2006-04-02 17:07:33 -0400621 if (unlikely((err = audit_ipc_obj(ipcp))))
622 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623 requested_mode = (flag >> 6) | (flag >> 3) | flag;
624 granted_mode = ipcp->mode;
625 if (current->euid == ipcp->cuid || current->euid == ipcp->uid)
626 granted_mode >>= 6;
627 else if (in_group_p(ipcp->cgid) || in_group_p(ipcp->gid))
628 granted_mode >>= 3;
629 /* is there some bit set in requested_mode but not in granted_mode? */
630 if ((requested_mode & ~granted_mode & 0007) &&
631 !capable(CAP_IPC_OWNER))
632 return -1;
633
634 return security_ipc_permission(ipcp, flag);
635}
636
637/*
638 * Functions to convert between the kern_ipc_perm structure and the
639 * old/new ipc_perm structures
640 */
641
642/**
643 * kernel_to_ipc64_perm - convert kernel ipc permissions to user
644 * @in: kernel permissions
645 * @out: new style IPC permissions
646 *
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800647 * Turn the kernel object @in into a set of permissions descriptions
648 * for returning to userspace (@out).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649 */
650
651
652void kernel_to_ipc64_perm (struct kern_ipc_perm *in, struct ipc64_perm *out)
653{
654 out->key = in->key;
655 out->uid = in->uid;
656 out->gid = in->gid;
657 out->cuid = in->cuid;
658 out->cgid = in->cgid;
659 out->mode = in->mode;
660 out->seq = in->seq;
661}
662
663/**
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700664 * ipc64_perm_to_ipc_perm - convert new ipc permissions to old
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665 * @in: new style IPC permissions
666 * @out: old style IPC permissions
667 *
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800668 * Turn the new style permissions object @in into a compatibility
669 * object and store it into the @out pointer.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670 */
671
672void ipc64_perm_to_ipc_perm (struct ipc64_perm *in, struct ipc_perm *out)
673{
674 out->key = in->key;
675 SET_UID(out->uid, in->uid);
676 SET_GID(out->gid, in->gid);
677 SET_UID(out->cuid, in->cuid);
678 SET_GID(out->cgid, in->cgid);
679 out->mode = in->mode;
680 out->seq = in->seq;
681}
682
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700683/**
Nadia Derbey3e148c72007-10-18 23:40:54 -0700684 * ipc_lock - Lock an ipc structure without rw_mutex held
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700685 * @ids: IPC identifier set
686 * @id: ipc id to look for
687 *
688 * Look for an id in the ipc ids idr and lock the associated ipc object.
689 *
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700690 * The ipc object is locked on exit.
691 */
692
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700693struct kern_ipc_perm *ipc_lock(struct ipc_ids *ids, int id)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700695 struct kern_ipc_perm *out;
Nadia Derbeyce621f52007-10-18 23:40:52 -0700696 int lid = ipcid_to_idx(id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697
698 rcu_read_lock();
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700699 out = idr_find(&ids->ipcs_idr, lid);
700 if (out == NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701 rcu_read_unlock();
Nadia Derbey023a5352007-10-18 23:40:51 -0700702 return ERR_PTR(-EINVAL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703 }
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700704
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705 spin_lock(&out->lock);
706
707 /* ipc_rmid() may have already freed the ID while ipc_lock
708 * was spinning: here verify that the structure is still valid
709 */
710 if (out->deleted) {
711 spin_unlock(&out->lock);
712 rcu_read_unlock();
Nadia Derbey023a5352007-10-18 23:40:51 -0700713 return ERR_PTR(-EINVAL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714 }
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700715
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 return out;
717}
718
Pavel Emelyanovb2d75cd2008-02-08 04:18:54 -0800719struct kern_ipc_perm *ipc_lock_check(struct ipc_ids *ids, int id)
720{
721 struct kern_ipc_perm *out;
722
723 out = ipc_lock(ids, id);
724 if (IS_ERR(out))
725 return out;
726
727 if (ipc_checkid(out, id)) {
728 ipc_unlock(out);
729 return ERR_PTR(-EIDRM);
730 }
731
732 return out;
733}
734
735/**
736 * ipcget - Common sys_*get() code
737 * @ns : namsepace
738 * @ids : IPC identifier set
739 * @ops : operations to be called on ipc object creation, permission checks
740 * and further checks
741 * @params : the parameters needed by the previous operations.
742 *
743 * Common routine called by sys_msgget(), sys_semget() and sys_shmget().
744 */
745int ipcget(struct ipc_namespace *ns, struct ipc_ids *ids,
746 struct ipc_ops *ops, struct ipc_params *params)
747{
748 if (params->key == IPC_PRIVATE)
749 return ipcget_new(ns, ids, ops, params);
750 else
751 return ipcget_public(ns, ids, ops, params);
752}
753
Pierre Peiffer8f4a3802008-04-29 01:00:51 -0700754/**
755 * ipc_update_perm - update the permissions of an IPC.
756 * @in: the permission given as input.
757 * @out: the permission of the ipc to set.
758 */
759void ipc_update_perm(struct ipc64_perm *in, struct kern_ipc_perm *out)
760{
761 out->uid = in->uid;
762 out->gid = in->gid;
763 out->mode = (out->mode & ~S_IRWXUGO)
764 | (in->mode & S_IRWXUGO);
765}
766
Pierre Peiffera5f75e72008-04-29 01:00:54 -0700767/**
768 * ipcctl_pre_down - retrieve an ipc and check permissions for some IPC_XXX cmd
769 * @ids: the table of ids where to look for the ipc
770 * @id: the id of the ipc to retrieve
771 * @cmd: the cmd to check
772 * @perm: the permission to set
773 * @extra_perm: one extra permission parameter used by msq
774 *
775 * This function does some common audit and permissions check for some IPC_XXX
776 * cmd and is called from semctl_down, shmctl_down and msgctl_down.
777 * It must be called without any lock held and
778 * - retrieves the ipc with the given id in the given table.
779 * - performs some audit and permission check, depending on the given cmd
780 * - returns the ipc with both ipc and rw_mutex locks held in case of success
781 * or an err-code without any lock held otherwise.
782 */
783struct kern_ipc_perm *ipcctl_pre_down(struct ipc_ids *ids, int id, int cmd,
784 struct ipc64_perm *perm, int extra_perm)
785{
786 struct kern_ipc_perm *ipcp;
787 int err;
788
789 down_write(&ids->rw_mutex);
Nadia Derbey00c2bf82008-07-25 01:48:03 -0700790 ipcp = ipc_lock_check(ids, id);
Pierre Peiffera5f75e72008-04-29 01:00:54 -0700791 if (IS_ERR(ipcp)) {
792 err = PTR_ERR(ipcp);
793 goto out_up;
794 }
795
796 err = audit_ipc_obj(ipcp);
797 if (err)
798 goto out_unlock;
799
800 if (cmd == IPC_SET) {
801 err = audit_ipc_set_perm(extra_perm, perm->uid,
802 perm->gid, perm->mode);
803 if (err)
804 goto out_unlock;
805 }
806 if (current->euid == ipcp->cuid ||
807 current->euid == ipcp->uid || capable(CAP_SYS_ADMIN))
808 return ipcp;
809
810 err = -EPERM;
811out_unlock:
812 ipc_unlock(ipcp);
813out_up:
814 up_write(&ids->rw_mutex);
815 return ERR_PTR(err);
816}
817
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818#ifdef __ARCH_WANT_IPC_PARSE_VERSION
819
820
821/**
822 * ipc_parse_version - IPC call version
823 * @cmd: pointer to command
824 *
825 * Return IPC_64 for new style IPC and IPC_OLD for old style IPC.
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800826 * The @cmd value is turned from an encoding command and version into
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827 * just the command code.
828 */
829
830int ipc_parse_version (int *cmd)
831{
832 if (*cmd & IPC_64) {
833 *cmd ^= IPC_64;
834 return IPC_64;
835 } else {
836 return IPC_OLD;
837 }
838}
839
840#endif /* __ARCH_WANT_IPC_PARSE_VERSION */
Mike Waychisonae781772005-09-06 15:17:09 -0700841
842#ifdef CONFIG_PROC_FS
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800843struct ipc_proc_iter {
844 struct ipc_namespace *ns;
845 struct ipc_proc_iface *iface;
846};
847
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700848/*
849 * This routine locks the ipc structure found at least at position pos.
850 */
Adrian Bunkb524b9a2008-02-06 01:36:28 -0800851static struct kern_ipc_perm *sysvipc_find_ipc(struct ipc_ids *ids, loff_t pos,
852 loff_t *new_pos)
Mike Waychisonae781772005-09-06 15:17:09 -0700853{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700854 struct kern_ipc_perm *ipc;
855 int total, id;
Kirill Korotaev73ea4132006-10-02 02:18:20 -0700856
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700857 total = 0;
858 for (id = 0; id < pos && total < ids->in_use; id++) {
859 ipc = idr_find(&ids->ipcs_idr, id);
860 if (ipc != NULL)
861 total++;
862 }
Mike Waychisonae781772005-09-06 15:17:09 -0700863
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700864 if (total >= ids->in_use)
865 return NULL;
Mike Waychisonae781772005-09-06 15:17:09 -0700866
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700867 for ( ; pos < IPCMNI; pos++) {
868 ipc = idr_find(&ids->ipcs_idr, pos);
869 if (ipc != NULL) {
870 *new_pos = pos + 1;
871 ipc_lock_by_ptr(ipc);
Mike Waychisonae781772005-09-06 15:17:09 -0700872 return ipc;
873 }
874 }
875
876 /* Out of range - return NULL to terminate iteration */
877 return NULL;
878}
879
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700880static void *sysvipc_proc_next(struct seq_file *s, void *it, loff_t *pos)
881{
882 struct ipc_proc_iter *iter = s->private;
883 struct ipc_proc_iface *iface = iter->iface;
884 struct kern_ipc_perm *ipc = it;
885
886 /* If we had an ipc id locked before, unlock it */
887 if (ipc && ipc != SEQ_START_TOKEN)
888 ipc_unlock(ipc);
889
Pierre Peiffered2ddbf2008-02-08 04:18:57 -0800890 return sysvipc_find_ipc(&iter->ns->ids[iface->ids], *pos, pos);
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700891}
892
Mike Waychisonae781772005-09-06 15:17:09 -0700893/*
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700894 * File positions: pos 0 -> header, pos n -> ipc id = n - 1.
895 * SeqFile iterator: iterator value locked ipc pointer or SEQ_TOKEN_START.
Mike Waychisonae781772005-09-06 15:17:09 -0700896 */
897static void *sysvipc_proc_start(struct seq_file *s, loff_t *pos)
898{
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800899 struct ipc_proc_iter *iter = s->private;
900 struct ipc_proc_iface *iface = iter->iface;
Kirill Korotaev73ea4132006-10-02 02:18:20 -0700901 struct ipc_ids *ids;
902
Pierre Peiffered2ddbf2008-02-08 04:18:57 -0800903 ids = &iter->ns->ids[iface->ids];
Mike Waychisonae781772005-09-06 15:17:09 -0700904
905 /*
906 * Take the lock - this will be released by the corresponding
907 * call to stop().
908 */
Nadia Derbey3e148c72007-10-18 23:40:54 -0700909 down_read(&ids->rw_mutex);
Mike Waychisonae781772005-09-06 15:17:09 -0700910
911 /* pos < 0 is invalid */
912 if (*pos < 0)
913 return NULL;
914
915 /* pos == 0 means header */
916 if (*pos == 0)
917 return SEQ_START_TOKEN;
918
919 /* Find the (pos-1)th ipc */
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700920 return sysvipc_find_ipc(ids, *pos - 1, pos);
Mike Waychisonae781772005-09-06 15:17:09 -0700921}
922
923static void sysvipc_proc_stop(struct seq_file *s, void *it)
924{
925 struct kern_ipc_perm *ipc = it;
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800926 struct ipc_proc_iter *iter = s->private;
927 struct ipc_proc_iface *iface = iter->iface;
Kirill Korotaev73ea4132006-10-02 02:18:20 -0700928 struct ipc_ids *ids;
Mike Waychisonae781772005-09-06 15:17:09 -0700929
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700930 /* If we had a locked structure, release it */
Mike Waychisonae781772005-09-06 15:17:09 -0700931 if (ipc && ipc != SEQ_START_TOKEN)
932 ipc_unlock(ipc);
933
Pierre Peiffered2ddbf2008-02-08 04:18:57 -0800934 ids = &iter->ns->ids[iface->ids];
Mike Waychisonae781772005-09-06 15:17:09 -0700935 /* Release the lock we took in start() */
Nadia Derbey3e148c72007-10-18 23:40:54 -0700936 up_read(&ids->rw_mutex);
Mike Waychisonae781772005-09-06 15:17:09 -0700937}
938
939static int sysvipc_proc_show(struct seq_file *s, void *it)
940{
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800941 struct ipc_proc_iter *iter = s->private;
942 struct ipc_proc_iface *iface = iter->iface;
Mike Waychisonae781772005-09-06 15:17:09 -0700943
944 if (it == SEQ_START_TOKEN)
945 return seq_puts(s, iface->header);
946
947 return iface->show(s, it);
948}
949
950static struct seq_operations sysvipc_proc_seqops = {
951 .start = sysvipc_proc_start,
952 .stop = sysvipc_proc_stop,
953 .next = sysvipc_proc_next,
954 .show = sysvipc_proc_show,
955};
956
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800957static int sysvipc_proc_open(struct inode *inode, struct file *file)
958{
Mike Waychisonae781772005-09-06 15:17:09 -0700959 int ret;
960 struct seq_file *seq;
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800961 struct ipc_proc_iter *iter;
962
963 ret = -ENOMEM;
964 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
965 if (!iter)
966 goto out;
Mike Waychisonae781772005-09-06 15:17:09 -0700967
968 ret = seq_open(file, &sysvipc_proc_seqops);
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800969 if (ret)
970 goto out_kfree;
971
972 seq = file->private_data;
973 seq->private = iter;
974
975 iter->iface = PDE(inode)->data;
976 iter->ns = get_ipc_ns(current->nsproxy->ipc_ns);
977out:
Mike Waychisonae781772005-09-06 15:17:09 -0700978 return ret;
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800979out_kfree:
980 kfree(iter);
981 goto out;
982}
983
984static int sysvipc_proc_release(struct inode *inode, struct file *file)
985{
986 struct seq_file *seq = file->private_data;
987 struct ipc_proc_iter *iter = seq->private;
988 put_ipc_ns(iter->ns);
989 return seq_release_private(inode, file);
Mike Waychisonae781772005-09-06 15:17:09 -0700990}
991
Arjan van de Ven9a321442007-02-12 00:55:35 -0800992static const struct file_operations sysvipc_proc_fops = {
Mike Waychisonae781772005-09-06 15:17:09 -0700993 .open = sysvipc_proc_open,
994 .read = seq_read,
995 .llseek = seq_lseek,
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800996 .release = sysvipc_proc_release,
Mike Waychisonae781772005-09-06 15:17:09 -0700997};
998#endif /* CONFIG_PROC_FS */